| There are a lot of optimization problems in the real world,which are generally divided into explicit objective optimization problems and implicit objective optimization problems.The optimization objectives of the former can be expressed by specific mathematical functions,and evolutionary algorithms have great advantages in solving such problems.However,the optimization objectives for implicit objective optimization problems such as indoor layout and music creation cannot be expressed by explicit mathematical functions,and evolutionary algorithms are typically difficult to solve.This type of challenge yields an interactive evolutionary algorithm.By introducing user-subjective preferences into the solution generation process,human subjective initiative and information processing capabilities are fully exploited.The interactive differential evolutionary algorithm,as one of its implementation methods,is an effective method for solving implicit objective optimization problems.It has been applied in many fields and has attracted the attention of many scholars at home and abroad.However,the IDE requires human intervention in evaluating decisions,which requires time.If there are too many individuals to evaluate,it is easy to create user fatigue,which limits how extensively the algorithm may be used and marketed.To address the issue of user fatigue in the IDE,the research state of interactive differential evolution and differential evolution is discussed and summarized in this thesis.Starting from the causes of user fatigue,a proxy evaluation model is introduced to replace user evaluation to a certain extent,and a convergence acceleration strategy suitable for IDE algorithm is proposed.Applied to pencil drawing generation and cartoon facial image design,respectively,and verified the effectiveness of the algorithms.It manifests itself specifically in the following ways:(1)An interactive differential evolution algorithm based on SVM is proposed that uses SVM as an agent model to realize intelligent evaluation.Based on the evaluation method of IDE pairwise comparison,the evaluation of users is regarded as a binary classification problem,and SVM is used as a classifier to implement proxy evaluation,and its classification accuracy is improved using particle swarm optimization.The whole evolution process is divided into three stages: the first,the middle,and the last.At the early stage of evolution,the user is in good condition,and SVM does not have enough training data,so the evaluation is directly completed by the user.In the middle of the evolution,users are tired.SVM is used as the agent model to complete the evaluation.In the later stage of evolution,the user state is adjusted,and the agent model and user interaction are used for evaluation.The method is applied to pencil drawing generation and compared with other methods.The results of the experiments show that the proposed method reduces user fatigue to some extent and that the IDE is feasible for pencil drawing.(2)A dual-opposition learning interactive differential evolution algorithm is proposed.First,use Tent mapping and opposition-based learning to improve the initial population and enhance population diversity.Secondly,the introduction of dual opposition-based learning can reasonably select opposition-based learning or centroid-based opposing learning based on population diversity to avoid excessive exploration or development that may result from a single opposing learning strategy.Finally,establish a protection mechanism to achieve a balance between exploration and development.Based on this method,a cartoon face generation system is constructed,and its performance is verified by comparing it with IDEs and IDE variants that apply single opposition learning.(3)The user’s dynamic selection strategy is introduced into the interactive differential evolution algorithm.The user’s dynamic selection means allowing users to modify each trait of an evolved individual according to their emotional preferences to meet their individual needs.This strategy can be incorporated into the IDE and its variants as a generic framework.The algorithm based on user dynamic selection is applied to cartoon face generation,and experiments are compared with the original algorithm.The experimental results show that the method integrated into the strategy has more advantages in reducing the number of iterations and meeting the personalized needs of users. |
